Atlas

Compare Models

Gemini 4 Argon vs Inkling 2 Small

Gemini 4 Argon leads by 5.8 capability points (≈95% interval +3.5 to +8.1): a clear lead over Inkling 2 Small · P(higher) ≈ 100%

Wins 82 of 108 shared benchmarks (20 lost, 6 tied)

Capability vs. relative cost
Gemini 4 ArgonInkling 2 SmallClosed weightsOpen weights

Each line runs through one model’s priced reasoning settings, cheapest to most expensive; grey points are other models’ priced settings in the same range. Results whose reasoning effort the source didn’t report are left out wherever a model has named effort levels. Capability and relative cost are the published estimates (relative cost is cost per task as a multiple of GPT-5 (high)); hover a setting, or use the arrow keys, to set it beside the other model’s same setting.

Gemini 4 ArgonvsInkling 2 Smallcapability vs. relative costData Nov 15, 2026Village LLM Atlasatlas.village.dev

Benchmark wins82 – 20 · 108 shared, 6 tied

Comparing matching setups only (53 benchmarks excluded) · suites counted once
← Gemini 4 ArgonInkling 2 Small →

Gemini 4 ArgonvsInkling 2 Smallwins by categoryData Nov 15, 2026Data: Vals AI, Artificial Analysis, MathArena and 12 othersVillage LLM Atlasatlas.village.dev

Margins

Each benchmark scaled 0–100% (its declared range, else the field’s range); Δ is Inkling 2 Small minus Gemini 4 Argon.

ArXivMath 09/2026 (Native Harnesses)Math87.731.1−56.6
ArXivMath 10/2026 (Native Harnesses)Math83.027.7−55.3
ArXivMath 08/2026 (Native Harnesses)Math82.536.8−45.6
Terminal-Bench 4.0Code57.612.6−45.0
Terminal-Bench-Science 0.1Science48.65.7−42.9
SRE Bench — Fully SolvedCode44.31.9−42.4
Terminal-Bench 4.0 (Artificial Analysis)Code55.113.1−41.9
Terminal-Bench Science 0.1 (Vals)Science44.32.9−41.4
CrossSearch (pass^3)Search64.023.0−41.0
Code Migration v1.1 (Vals Index v2.2 subset)Code74.439.5−35.0
ChartographyMultimodal71.636.7−34.9
HLAB-Vals-1.1 - Arbitration International Dispute ResolutionLegal0.030.0+30.0
HLAB-Vals-1.1 - Banking FinanceLegal40.010.0−30.0
HLAB-Vals-1.1 - Corporate GovernanceLegal60.030.0−30.0
HLAB-Vals-1.1 - Data Privacy CybersecurityLegal50.020.0−30.0
HLAB-Vals-1.1 - Funds Asset ManagementLegal60.030.0−30.0
HLAB-Vals-1.1 - Healthcare Life SciencesLegal60.030.0−30.0
HLAB-Vals-1.1 - ImmigrationLegal0.030.0+30.0
HLAB-Vals-1.1 - Litigation Dispute ResolutionLegal10.040.0+30.0
Legal Research Bench ImmigrationLegal54.527.3−27.3
APEX-Agents 1.1Professional Work87.460.9−26.5
Terminal-Bench 2.1Code57.781.3+23.6
BrokenArXiv 10/2026 (Native Harnesses)Math39.863.3+23.5
Vals Science IndexIndexes55.432.7−22.7
Humanity's Sixth SenseMultimodal55.433.3−22.1
MysteryMechanismScience45.525.2−20.3
HLAB-Vals-1.1 - Antitrust CompetitionLegal60.040.0−20.0
HLAB-Vals-1.1 - Bankruptcy RestructuringLegal0.020.0+20.0
HLAB-Vals-1.1 - Capital MarketsLegal40.020.0−20.0
HLAB-Vals-1.1 - Environmental ESGLegal20.00.0−20.0
HLAB-Vals-1.1 - InsuranceLegal20.00.0−20.0
HLAB-Vals-1.1 - Intellectual PropertyLegal40.020.0−20.0
HLAB-Vals-1.1 - Real EstateLegal10.030.0+20.0
Legal Research Bench CriminalLegal56.136.8−19.3
HANDBOOK.md Agents v1.1Professional Work36.617.8−18.8
AA-3D v1.0 EloCode11471027−17.3
CritPt v2 (agentic, avg@5)Science72.955.7−17.1
VIALS-AA (pass^3)Multimodal59.642.9−16.8
DAYJOB: Finance (October 2026 rescoring)Finance & Accounting30.514.3−16.2
AA-Omniscience v1.1 IndexIndexes46.116.1−15.0
Humanity's Last Exam DiamondGeneral QA42.528.7−13.8
Legal Research Bench FamilyLegal36.422.7−13.6
BrokenArXiv 09/2026 (Native Harnesses)Math49.763.2+13.5
Tax Agent BenchFinance & Accounting79.065.8−13.3
SWE-bench Multimodal (Vals v1, 480 Tasks)Code56.043.1−12.9
Finance Agent v2.1 - Market AnalysisFinance & Accounting70.759.0−11.7
DeepFinanceFinance & Accounting41.429.8−11.6
Excel Modeling Benchmark v1.1 LBOFinance & Accounting67.255.9−11.3
HLAB-Vals-1.1 - Corporate M&ALegal30.020.0−10.0
HLAB-Vals-1.1 - Energy Natural ResourcesLegal40.050.0+10.0
HLAB-Vals-1.1 - International Trade SanctionsLegal10.020.0+10.0
HLAB-Vals-1.1 - Structured Finance SecuritizationLegal30.020.0−10.0
HLAB-Vals-1.1 - Trusts Estates Private ClientLegal30.020.0−10.0
Excel Modeling Benchmark v1.1 CompsFinance & Accounting82.973.1−9.8
ARC-AGI-2General QA91.381.7−9.6
APEX-Accounting 1.1Finance & Accounting71.562.0−9.5
SciCode 1.0.1 (Artificial Analysis)Code61.852.3−9.5
BrokenArXiv 08/2026 (Native Harnesses)Math54.263.1+8.9
ComplexConstraintsChat & Writing53.344.5−8.8
VIALS-AA (average of 3)Multimodal66.357.8−8.5
SWE Atlas v1.1 - OverallCode70.161.9−8.2
Excel Modeling Benchmark v1.1 TemplateFinance & Accounting79.671.6−8.0
Legal Research Bench Constitutional/Civil RightsLegal47.439.5−7.9
Legal Research Bench Administrative/RegulatoryLegal68.260.6−7.6
SAGE v1.1 - Partial Differential EquationsMath87.094.0+7.0
Finance Agent v2.1 - PrecedentsFinance & Accounting40.447.1+6.7
Legal Research Bench HealthLegal76.770.0−6.7
Medical Long Context Reasoning (MLCR-AA, October 2026 methodology)Health & Medicine13.920.6+6.7
Excel Modeling Benchmark v1.1 M&AFinance & Accounting80.273.7−6.6
Finance Agent v2.1 - Earnings AnalysisFinance & Accounting82.688.9+6.3
BioMysteryBench (Vals Terminus 2)Science76.370.0−6.3
Arena Code WebDev — ReactCode17041659−6.0
Finance Agent v2.1 - Financial ModelingFinance & Accounting36.542.4+5.9
FrontierCode 2.0 - Pass RateCode30.625.0−5.6
FrontierCode 2.0 - ScoreCode27.622.3−5.4
CyberBench v1.1 OverallCybersecurity77.972.7−5.1
SWE-Marathon v1.1Code43.838.8−5.0
Excel Modeling Benchmark v1.1 MiscFinance & Accounting73.068.3−4.8
Finance Agent v2.1 - General Quantitative AnalysisFinance & Accounting83.779.2−4.5
FrontierMemory v1.1Agents64.460.0−4.4
SAGE v1.1 - Computer ScienceMath95.091.2−3.7
Finance Agent v2.1 - AdjustmentsFinance & Accounting70.867.4−3.4
Excel Modeling Benchmark v1.1 DCFFinance & Accounting65.562.2−3.3
Finance Agent v2.1 - ComparablesFinance & Accounting61.765.0+3.3
Legal Research Bench Weighted PassLegal91.087.9−3.1
Excel Modeling Benchmark v1.1 ScratchFinance & Accounting70.067.1−2.9
HLAB-Vals-1.1 - Criteria Pass RateLegal96.393.7−2.6
Finance Agent v2.1 - Disclosure AnalysisFinance & Accounting83.380.8−2.6
AA-AnalystAgent (pass^5)Professional Work48.851.3+2.5
AA-Briefcase v1.1 EloAgents14851527+2.3
Finance Agent v2.1 - All PassFinance & Accounting58.956.6−2.3
sudo L7 (October 2026 rescoring)Code16.114.1−2.0
ARC-AGI-1 Semi-PrivateGeneral QA98.596.5−2.0
ProgramBenchCode2.50.5−2.0
Excel Modeling Benchmark v1.1 DataroomFinance & Accounting85.083.4−1.6
SAGE v1.1 - Linear AlgebraMath81.980.5−1.4
Finance Agent v2.1 - General Qualitative AnalysisFinance & Accounting79.278.0−1.1
VIALS-AA (pass@3)Multimodal72.772.0−0.6
GDP.pdf (Artificial Analysis, All-pass)Professional Work25.625.2−0.4
GDPval-AA v2.1Professional Work16341630−0.1
Excel Modeling Benchmark v1.1 OperatingFinance & Accounting69.970.0+0.1
Arena Code WebDev — HTMLCode166016600.0
HLAB-Vals-1.1 - Emerging Companies Venture CapitalLegal20.020.00.0
HLAB-Vals-1.1 - Employment LaborLegal0.00.00.0
HLAB-Vals-1.1 - TaxLegal20.020.00.0
HLAB-Vals-1.1 - White Collar Defense InvestigationsLegal0.00.00.0
Legal Research Bench Business/CommercialLegal46.546.50.0
Legal Research Bench Civil LitigationLegal51.951.90.0

Gemini 4 ArgonvsInkling 2 Smallnormalised score marginsData Nov 15, 2026Data: Vals AI, Artificial Analysis, MathArena and 12 othersVillage LLM Atlasatlas.village.dev

Win counts use each benchmark’s declared score direction. Represented suite slices and the derived Capability score are excluded from headline totals. Missing or incompatible setups are excluded, never counted as losses.
Loading Atlas data…